Cisco launches a new AI data center chip that reduces energy expenditure by 70%

New York. – The American technology company Cisco presented this Tuesday its new Silicon One G300 chip, a high-capacity network architecture designed to accelerate processing in massive artificial intelligence (AI) data centers and reduce energy consumption by up to 70% through liquid cooling systems, which seeks to compete with the offerings of Broadcom and Nvidia. Within the framework of the Cisco Live EMEA event held in Amsterdam, the company detailed that this new silicon achieves a switching capacity of 102.4 Terabits per second (Tbps), allowing the management of AI clusters at gigawatt scale for training and inference tasks.

We recommend reading: Brussels threatens Meta with measures if it prevents competition in WhatsApp AI

“We are building the foundations of the infrastructure of the future,” said Jeetu Patel, President and Chief Product Officer of Cisco, who emphasized that innovation spans from silicon to software to support the deployment of the “era of AI agents.” The G300 chip, manufactured with 3-nanometer technology, includes proactive telemetry functions to prevent data packet loss and optimize the use of graphics processing units (GPUs), thereby maximizing the return on investment in computing for its customers. The company confirmed that both the G300 silicon and the new advanced optical systems will begin to be distributed throughout this year, with the support of strategic partners such as Intel, AMD and Nvidia.